Minimum (ir maximum) bam depth/coverage for structural variation analyses

[linudz]

*or maximum

We are studying structural variation in several plant genomes Illmumina pair-end DNA re-sequencing, and noticed that the number of deletions in some samples (PINDEL) drops significantly. After having boldly calculated the depth as

[number of reads] x mean read length / genome size

we noticed that in these samples the depth was way lower (around 10 folds on an average of 50).

I was wondering if there is some lower or higher limit in a sample depth to start considering an adulterated result.

Thanks.